[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] RE: [Xen-users] (Xenlinux) gentoo-xen-kernel patches questions?


  • To: Mike Viau <viaum@xxxxxxxxxxxxxxx>
  • From: Andrew Lyon <andrew.lyon@xxxxxxxxx>
  • Date: Sun, 4 Apr 2010 13:59:41 +0100
  • Cc: bderzhavets@xxxxxxxxx, xen-devel@xxxxxxxxxxxxxxxxxxx, xen-users@xxxxxxxxxxxxxxxxxxx
  • Delivery-date: Sun, 04 Apr 2010 06:01:06 -0700
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type; b=qD6qwZDrpw1seXSGkJS0jTcYPWtkVtlwb2wlnquIxSnh4ndcIdQk8iSEtgV6dD1JPo 6Sg1N1fIoGyZuQQAa/uL7x4j2TJIeilHfsmVRC3RuAeQrziaQqAv6t/CFmy6XhAr0MI4 ct2Ep6bBe6JBepg3Tn+K4sc1bhX/FKYo2DhQc=
  • List-id: Xen user discussion <xen-users.lists.xensource.com>



On Sat, Apr 3, 2010 at 3:49 PM, Mike Viau <viaum@xxxxxxxxxxxxxxx> wrote:
On Sat, Apr 3, 2010 at 4:11 AM, Mike Viau <viaum@xxxxxxxxxxxxxxx> wrote:
Fri, 2 Apr 2010 17:21:35 -0700 <bderzhavets@xxxxxxxxx> wrote:

>>Would compiling the pv_ops kernel from the git repository produce any additional stability?

>Good question. It completely depends on speed of your internet connection
and :-
># git checkout -b xen/linux-2.6.32.X-stable  origin/xen/linux-2.6.32.X-stable
># make menuconfig ( to enable Xen Dom0 support)
># make -j(x)number_of_cores
># make modules_install install
># mkinitramfs -o /boot/initrd-2.6.32.10.img 2.6.32.10

Yes I was in my plans to compare the debian pvops kernel against a pvops compiled kernel from Jeremy's git repo.

Additionally I hoped to acquire a 2.6.33 Xenified kernel .config to compile a forward ported 2.6.33 Xenlinux kernel for testing with.


__________________________________________________________________________


Sat, 3 Apr 2010 15:17:47 +0100 <andrew.lyon@xxxxxxxxx> wrote:

>Hi Mike,
>
>I depreciated the 2.6.33 patches because they turned out to be very unstable, I've left 2.6.31 because it is very much stable, in fact in the next few days I have a updated 2.6.31 patchset and ebuild to release which I hope to get unmasked in Gentoo portage so that it is just keyword ~ rather than being masked out as it currently is.


That sounds great for xen users that still use 3rd party kernel modules. By the way I noticed the Xenified kernels worked in Xen 3.4.x, but I got a boot load error about not being able to read the file (likely the vmlinuz) with Xen 4.0.0. Is this normal because 4.0.0 is using the pv_ops as the default kernel?

I've not tried Xenified kernel + 4.0 for a while but I did try xen-4.0.0-rc4-pre a while ago and it worked ok (still have the boot logs) so I think perhaps you had the wrong grub syntax or there is a new bug since I last tried it, I will give it a try next week.
 

If this is the case I plan to test stability against pv_ops & Xenlinux kernels with Xen 3.4.3 (which seems to work with both kernel types) when it gets finished and released. Am I losing any features by going back to 3.4.3 or even the older 3.4.2 release?

I think 3.4.3 or 3.4.2 would be a very good choice for testing stability and performance of the two different dom0 flavours, its certainly the version I would use in production, 4.x has lots of new features but it is very new and I wouldn't be putting it into production use just yet.
 



>I've also got a 2.6.32 ebuild which is stable so far on all of my Xen systems, provided there are no problems in the next few days I will release that as well, both .31 and .32 will be maintained for a while as they are the official kernels for openSUSE 11.2 and SLE 11 SP1.


That should put the Xenlinux kernel versions inline with the pv_ops ones.



>While there are branches in the openSUSE/Novell git tree for openSUSE 11.2 and SLE11-SP1 the master tree tracks kernel.org and has already passed 2.6.33 so there doesn't seem much value in trying to make a Xenlinux forward ported 2.6.33, I might do .34 when it is released but I tend to wait until they branch for a distro release so that there is long-term support.


Makes perfect sense. For your information, despite the fact that I didn't like my minimalistic kernel configuration, I was able to run a xenified 2.6.33 kernel for days with high load (on Xen 3.4.3-rc3) and it stayed up for me.



>I release these patches because people have asked me to, and I make them because pv_ops is not fully functional yet and doesn't work with all of the 3rd party kernel modules etc that I have to use, for example I run Xen on 3 very powerful workstations and all of them have nvidia and acronis snapapi drivers loaded.


I personally do really appreciate it :) Andy


-M



My current 2.6.33 .config is attached. I was hoping to have acquired a more generic config rather than just targeting primarily for my system. I do not mind if it will take longer to build the kernel :-)

Thanks.


-M


--- On Fri, 4/2/10, Mike Viau <viaum@xxxxxxxxxxxxxxx> wrote:

>>Hi,
>>
>>I noticed that the Xenlinux 2.6.33 patches are tagged as being "Deprecated" but the 2.6.31 are not.
>>
>>May I ask why?
>>
>>http://code.google.com/p/gentoo-xen-kernel/downloads/list?can=1&q=&colspec=Filename+Summary+Uploaded+Size+DownloadCount
>>
>>
> Fri, 2 Apr 2010 13:15:09 -0400 <konrad.wilk@xxxxxxxxxx> wrote:
>
> No idea. But those patches are pursuing an entirely different method of
> getting Linux to work in privileged state (Dom0). The current
> development is with the pv-ops one:
>
> http://wiki.xensource.com/xenwiki/XenParavirtOps
>

I am currently using the unstable version of debain's pvops xen kernel.

http://packages.debian.org/sid/linux-image-2.6.32-4-xen-amd64

It appears to have functioning networking and block devices which is nice, but I have experienced some instabilities such as kernel dumping on reboots. Unfortunately I do not have a console setup to log kernel messages for this machine.

Would compiling the pv_ops kernel from the git repository produce any additional stability?

It was my understanding that one could run a stable Xen systems today with a Xenlinux kernel, as pv_ops was still in development. I've read on the XenParavirtsOps page that there are features available to Xen but are not yet available in the pv_ops kernel (perhaps this wiki has not been updated as I understand Xen 4.0 release is just around the corner!):

 
  • Work in progress:
    • dom0 support, currently planned for Linux 2.6.35 (latest pv_ops dom0 patches can be found from jeremy's git tree, see instructions below)
    • pv-on-hvm driver support
    • Balloon expansion (using memory hotplug) to grow bigger than initial domU memory size
  • To be done:
    • Device hotplug
    • Other device drivers
    • kdump/kexec
    • pvscsi backend (dom0)
    • pvusb backend (dom0)
    • ...?


>>Also is there somewhere one could view list xen features added into linux kernel source (e.g: netchannel2, blktap, etc) ?
>>

Does anyone have a graph or something showing the state of these Xen features/modules in the two predominate Dom 0 capable kernels (Xenlinux & PV_OPS) to share or post to a wiki?

root@localhost:~# tree /lib/modules/2.6.32-4-xen-amd64/kernel/drivers/xen
/lib/modules/2.6.32-4-xen-amd64/kernel/drivers/xen
|-- blktap
|   `-- blktap.ko
|-- xen-evtchn.ko
`-- xenfs
    `-- xenfs.ko

2 directories, 3 files

Other XEN features are compiled as (=y) in the unstable version of debain's pvops xen kernel .config file.

>>
>>Lastly could someone provide perhaps a few xenlinux kernel .configs that works. I have attached my working config used with 2.6.33 source, but I am hoping to get a copy of a more generic + working .config
>>

Recently pv_ops kernel configs for both the 2.6.31 and 2.6.32 kernels have been in circulation. Can anyone share some .configs that were used on Xenlinux kernels (not 2.6.18) when they were in their prime?


Thanks!


-M
_______________________________________________
_______________________________________________

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.